The pool of memory-phenotype CD8 T cells is composed of Ag-induced (AI) and cytokine-induced innate (IN) cells. IN cells have been described as having properties similar to those of AI memory cells. However, we found that pathogen-induced AI memory cells can be distinguished in mice from naturally generated IN memory cells by surface expression of NKG2D. Using this marker, we described the increased functionalities of AI and IN memory CD8 T cells compared with naive cells, as shown by comprehensive analysis of cytokine secretion and gene expression. However, AI differed from IN memory CD8 T cells by their capacity to migrate to the lung parenchyma upon inflammation or infection, a process dependent on their expression of ITGA1/CD49a and ITGA4/CD49d integrins.

One hallmark of the adaptive immune system is its ability to respond more quickly and more strongly to previously encountered Ags. This immunological memory relies on the generation of cells that display increased reactivity toward the previously encountered Ag. Protection against intracellular pathogens or tumor-derived Ag is conferred in part by Ag-induced (AI) memory CD8 T cells. Indeed, AI memory CD8 T cells have improved functional properties compared with naive cells, making them more potent to rapidly eliminate infected cells upon reinfection (1, 2). These AI memory cells are found in secondary lymphoid organs, but a subset of them, the tissue-resident memory cells (TRM), settles within nonlymphoid tissues where they provide increased protection against secondary pathogen infections (3, 4). TRM are long-lived sessile cells in most tissues except the lung, where they need to be replenished from the circulating pool of memory cells (5, 6).

Memory-phenotype CD8 T cells or innate (IN) memory cells can also be generated through several alternative pathways that are independent of foreign Ag exposure (7). Memory CD8 T cells generated through lymphopenia-induced proliferation (LIP) were the first IN memory cells to be described (812). This pathway depends on strong IL-7 stimulation of naive CD8 T cells (because of the increased availability of this γc cytokine in the lymphopenic host) combined with weak TCR stimulation through self-peptide/MHC complexes (1315). Other γc cytokines also support the generation of IN memory cells. In vivo, strong IL-2 stimulation through injection of IL-2/anti–IL-2 Ab complexes was shown to drive the generation of IN memory CD8 T cells from naive TCR transgenic cells (16). Similarly, the characterization of several mutant mouse strains revealed that strong IL-4 stimulation of CD8 single-positive thymocytes or naive CD8 T cells leads to IN memory cell generation (1719). Moreover, naive BALB/c mice have an increased proportion of memory-phenotype CD8 T cells because of higher levels of circulating IL-4 compared with naive C57BL/6 mice (20). Conversely, naive mice deficient for IL-4 production or signaling have a reduced frequency of IN memory CD8 T cells (21, 22). In physiological conditions, LIP memory cells generation occurs during the neonatal period in naive mice (21, 23, 24) and Th2 immune responses might also favor the generation of IN memory CD8 T cells (25).

Hence, among CD8 T cells specific for foreign Ag never encountered by specific pathogen-free (SPF) naive mice, 10–20% of cells display a memory phenotype. These cells are also referred to as virtual memory (VM) cells (26). Importantly, equal numbers of these VM cells were found in naive germ-free mice, indicating that their generation is independent of microbiota-derived Ag (26). Therefore, in physiological conditions, the pool of memory-phenotype CD8 T cells is composed of two classes of cells: AI and IN.

AI and IN CD8 memory cells are generated through distinct pathways but express a similar array of surface markers, which has hampered their demarcation. Interestingly, experimentally generated TCR transgenic OT-I memory CD8 T cells do not express CD49d compared with AI OT-I memory CD8 T cells. This lack of CD49d expression has been used to identify and characterize OVA and vaccinia virus (VV)–specific clones of IN memory cells generated in physiological conditions (26). In parallel, it has been shown that compared with naive cells, these unconventional Ag-specific memory cells (IN or VM cells) are able to mount an efficient response against pathogen infection with increased functional properties including augmented IFN-γ production and proliferative response (18, 26, 27). However, the comparison between IN and AI memory CD8 T cells in terms of phenotype, function, and gene expression profile has not been performed.

Upon strong TCR triggering, CD8 T cells express high levels of the NK cell receptor NKG2D (28), and AI memory cells express NKG2D (29). In contrast, IL-4–induced IN cells do not express NKG2D (29). Therefore, we hypothesized that NKG2D could be differently expressed between AI and IN memory CD8 T cell populations.

In this study, we demonstrate that the expression of NKG2D is restricted to AI memory CD8 T cell populations. Using NKG2D as a marker of AI cells, we performed an extensive comparison of AI and IN cells within the natural pool of memory CD8 T cells. Our results indicate that although IN CD8 T cells share many features with AI memory cells, only AI cells are recruited toward the lung parenchyma upon inflammation or infection.

F5 TCR [B6/J-Tg(CD2-TcraF5,CD2-TcrbF5)1Kio/Jmar] transgenic mice were provided by Prof. D. Kioussis (National Institute of Medical Research, London, U.K.) and backcrossed on CD45.1 C57BL/6 background (30). The F5 TCR recognizes the NP68 peptide from influenza A virus (ASNENMDAM) in the context of H2-Db. OT-I TCR transgenic [B6/J-Tg(Tcra,Tcrb)1100Mjb/Crl], CD45.2 (C57BL/6J) and CD45.1 (B6.SJL-PtprcaPepcb/BoyCrl), C57BL/6J, BALB/c, and OF1 mice were purchased from Charles River Laboratories (L’Arbresle, France). The OT-I TCR recognizes the OVA257–264 peptide from chicken OVA (SIINFEKL) in the context of H2-Kb. Mice were bred or housed under SPF conditions in our animal facility (AniRA-PBES, Lyon, France). All experiments were approved by our local ethics committee (CECCAPP, Lyon, France) and accreditations have been obtained from governmental agencies.

The recombinant influenza (Flu) virus strain WSN encoding the NP68 epitope (Flu-NP68) was produced by reverse genetics from the A/WSN/33 H1N1 strain. The recombinant VV, expressing the NP68 epitope (VV-NP68), was engineered from the Western Reserve strain by Dr. D.Y.-L. Teoh, in Prof. Sir Andrew McMichael’s laboratory at the Medical Research Council (Human Immunology Unit, Institute of Molecular Medicine, Oxford, U.K.). The Listeria monocytogenes strain 10403s was produced from clones grown from organs of infected mice. For immunization, anesthetized mice received intranasal (i.n.) administration of Flu (2 × 105 tissue culture ID50), VV (2 × 105 PFU) or poly(I:C) (30 μg) in 20 μl of PBS or i.v. L. monocytogenes (2 × 103) administration in 200 μl PBS. For some indicated experiments, mice received i.p. administration of 1 × 106 PFU VV in 200 μl of PBS.

To generate AI TCR transgenic memory CD8 T cells, 2 × 105 naive CD45.1 F5 CD8 T cells were transferred in C57BL/6 mice by i.v. injection. The next day, mice were infected with VV-NP68 as described above. To generate IN TCR transgenic memory CD8 T cells, 1 × 106 naive OT-I CD8 T cells were transferred in sublethally irradiated (600 rad) CD45.1 C57BL/6 mice by i.v. injection. OT-I or F5 naive cells were also transferred to immunocompetent mice that further received i.p. injections of 1.5 μg of IL-2 or IL-4 (PeproTech) immunocomplexed to anti–IL-2 (S4B6; Bio X Cell) or anti-IL-4 (11B11; Bio X Cell) Ab, during seven consecutive days.

To discriminate between TRM and circulating memory CD8 T cells, in vivo intravascular staining was performed as previously described (31). Briefly, mice were injected i.v. with 3 μg of CD45-BV421 Ab (BioLegend) diluted in 200 μl of sterile PBS (Life Technologies, Saint Aubin, France) and were sacrificed 2 min after injection by overdose of pentobarbital. Blood samples (100 μl) were collected on EDTA by retro-orbital bleeding. Spleen and lymph nodes were harvested, mechanically disrupted, and filtered through a sterile 100-μm nylon mesh filter (BD Biosciences). To collect bronchoalveolar lavages, the trachea was exposed and cannulated with a 24-gauge plastic catheter (BD Biosciences) and lungs were lavaged twice with 1 ml of cold sterile PBS. Lungs were enzymatically digested using a specific dissociation kit and following manufacturer’s instructions (Miltenyi Biotec).

Surface staining was performed on single-cell suspensions from each organ for 30 min at 4°C with the appropriate mixture of mAbs diluted in staining buffer (PBS supplemented with 1% FCS [Life Technologies] and 0.09% NaN3 [Sigma-Aldrich, Saint Quentin-Fallavier, France]). To identify B8R-specific memory CD8 T cells, dextramer staining was performed for 20 min at room temperature using B8R dextramer (Immudex) before surface staining. The following Abs (clones) were used for surface staining: NKG2D (CX5), CD45.1 (A20), CD45.2 (104), CD122 (TM-b1), CD62L (MEL-14), CD8 (53-6.7), CD44 (IM7.8.1), CXCR3 (Cxcr3-173), CD49a (HA 31-8), CD49d (R1-2), CD29 (eBioHMb1-1), CD11c (N418). VM CD8 T cells were from nonimmunized naive C57BL/6 mice. Spleen CD8 T cells were enriched using a MACS CD8a+ T Cell Isolation Kit II for mouse (Miltenyi Biotec), and VM were identified as B8R-specific CD44hi CD49 CD8 T cells (using CD11b, CD11c, CD19 and NK1.1 as a dump gate). To perform intracellular cytokine staining, cells were fixed and permeabilized using CytoFix/CytoPerm (BD Pharmingen). To detect transcription factors, a Foxp3 Kit (eBioscience) was used to fix and permeabilize cells. The following Abs (clones) were used for intracellular staining: IFN-γ (XMG1.2), CCL5 (2E9), Tbet (4B10), and Eomes (Dan11mag). All analyses were performed on a BD Biosciences FACS LSR II or Fortessa and analyzed with FlowJo software (Tree Star, Ashland, OR).

For measurements of cytokine production, 5 × 104 naive and memory (NKG2D and NKG2D+) CD8 T cells sorted from VV-infected mice were cultured for 12 h with plate-bound anti-CD3 Ab (145-2C11, 10 μg/ml; BD Biosciences), soluble anti-CD28 Ab (37.51, 1 μg/ml; BD Biosciences), and IL-2 (2%). Supernatants were collected and cytokine production was measured by bead-based multiplexing technology for IL-1α, IL-1β, IL-3, IL-4, IL-9, IL-10, IL-13, IL-17, IFN-γ, TNF-α, and CCL2/3/4/5 (Bio-Plex Pro; Bio-Rad) or by ELISA for CCL1, CCL5, and IFN-γ (Mouse DuoSet; R&D Systems). For flow cytometry measurements of cytokine production at the single-cell level, 1 × 105 NKG2D and NKG2D+ memory CD8 T cells, sorted from VV-infected mice, were cultured for 6 h with VV-infected (multiplicity of infection = 10) DC2.4 cells or with PMA (20 ng/ml) and ionomycin (1 μg/ml) in the presence of GolgiStop (BD Biosciences). Alternatively, 1 × 106 total splenocytes from VV-infected mice were cultured for 5 h with plate bound anti-CD3 Ab and soluble anti-CD28 Ab or with IL-12 (10 ng/ml; R&D Systems), IL-18 (10 ng/ml; MBL), and IL-2 (10 ng/ml; PeproTech).

To evaluate the degree of protection associated with each CD8 T cells population, mice were transferred with 1 × 105 naive, NKG2D or NKG2D+ memory CD8 T cells from Flu-infected mice. The next day, host mice were infected with a lethal dose (1 × 106 tissue culture ID50) of Flu. Mice weight loss was measured each day, for up to 12 d, postinfection. Mice that lost more than 20% of initial body weight were euthanized.

Naive, NKG2D and NKG2D+ memory CD8 T cells were sorted from VV-infected mice (50 d postinfection). Cells were lysed and multiplex PCR were performed by ImmunID (Grenoble, France) on genomic DNA to detect V(D)J rearrangements at the TCR β-chain locus. For each cell population, the percentage of TCR repertoire diversity was calculated as the ratio between the number of observed V(D)J recombinations and the theoretical number of V(D)J recombinations (i.e., 209).

F5 TCR transgenic memory CD8 T cells were generated as described above. Eighty days after VV-NP68 infection, CD45.1 F5 memory CD8 T cells as well as host’s NKG2D and NKG2D+ memory CD8 T cells were sorted from five pools of spleens, each from eight mice. CD8 T cells were enriched by negative selection (MACS CD8a+ T Cell Isolation Kit II for mouse); then, memory CD8 T cell (CD8+CD44+) were sorted by FACS based on the expression of NKG2D (purity >98%). Naive F5 and polyclonal CD8 T cells were sorted from pools of three spleens from naive F5 and C57BL/6 mice respectively. Total RNA was extracted from dry cell pellets according to the “Purification of total RNA from animal and human cells” protocol of the RNeasy Micro Kit (Qiagen, Hilden, Germany). Purity and integrity of the RNA was assessed on the Agilent 2100 Bioanalyzer (Agilent, Palo Alto, CA). Total RNA from each sample was amplified, labeled, and hybridized to mouse GeneChip HT MG-430 PM Plates as described in the Affymetrix GeneChip 3′ IVT PLUS Reagent Kit User Manual (Affymetrix, Santa Clara, CA). Affymetrix CEL files were analyzed in R using the appropriate packages from the Bioconductor suite (https://www.bioconductor.org/). Raw probe signals were background corrected using the maximum likelihood estimation of the normal–exponential mixture model (32), normalized using the variance stabilization normalization (33) followed by a quantile normalization (34). Summarization was performed using the median (35) and using a modified version 17.1 of the Entrez-Gene–based reannotated chip description file. Noninformative genes were filtered using the I/NI algorithm (36). Linear models were applied using the limma package to compute the average expression level for each cell type. A random effect was introduced to account for the paired design. Statistical contrasts were then applied to compute differential expression between the different cell types. The empirical Bayes method was used to compute moderated p values that were then corrected for multiple comparisons using the Benjamini and Hochberg’s false discovery rate controlling procedure.

Naive, NKG2D, and NKG2D+ memory CD8 T cells were sorted from VV-infected mice (50 d postinfection). Total RNA was extracted using Trizol reagent according to manufacturer’s instructions (Life Technologies). Total RNA was digested using turbo DNA-free DNase (Life Technologies) to avoid genomic contamination. Quality and absence of genomic DNA contamination were assessed with a Bioanalyzer (Agilent, Massy, France). We used a High-Capacity RNA-to-cDNA Kit (Life Technologies) to generate cDNA for PCR amplification. PCR was carried out with a SYBR Green–based kit (FastStart Universal SYBR Green Master; Roche, Basel, Switzerland) on a StepOnePlus instrument (Applied Biosystems, Calrlsbad, CA). Primers were designed using the Roche Web site (Universal Probe Library Assay Design Center).

1 × 106 CD8 T cells purified from VV-infected mice (50 d postinfection) were added to the upper chamber of polycarbonate transwell inserts (5 μm pore size; Corning). The lower chamber was filled with complete DMEM medium (6% FCS, 10 mM HEPES, 50 μM 2−MΕ, 50 μg/ml gentamicin, 2 mM l-glutamine, all from Life Technologies) alone or supplemented with CXCL10 (100 ng; PeproTech). After 2 h of incubation at 37°C, 7% CO2, transmigrated cells were collected in the lower chamber and were washed and stained for CD8, CD44, and NKG2D. The absolute number of transmigrated cells was determined by flow cytometry by adding a known number of fluorescent beads (Flow-Count Fluorosphere; Beckman Coulter). Results are expressed as the migration index, which represents the fold increase in the number of transmigrated cells in response to chemoattractant over the nonspecific cell migration (medium alone).

P815 target cells were labeled with Cell Trace Violet (Thermo Fisher) and incubated with anti-CD3 Ab (2C11, 10 μg/ml) for 30 min at 37°C. Target cells were cultured for 12 h with naive, NKG2D, or NKG2D+ memory CD8 T cells sorted from VV-infected C57BL/6 mice (80 d postinfection). P815 viability was assessed using LIVE/DEAD fixable dye (Life Technologies) followed by fixation with 1% paraformaldehyde (Sigma-Aldrich) followed by flow cytometry analysis. The percentage of specific lysis is the percentage of P815 cell death in samples containing CD8 T cells after subtracting the percentage of spontaneous P815 cell death.

We tested the hypothesis that NKG2D expression could discriminate AI from IN memory cells. Indeed, IN cells generated following injection of IL-2 or IL-4 Ab complexes (Fig. 1A, Supplemental Fig. 1) or after lymphopenic proliferation (Supplemental Fig. 2) do not express NKG2D at their cell surface. In agreement with this hypothesis, in nonimmunized SPF mice, the majority (more than 90%) of splenic memory-phenotype CD8 T cells (i.e., CD44hi), which are mainly IN cells (7), do not express NKG2D (Fig. 1B). Similarly, VM cells are mainly NKG2D (Fig. 1C). In contrast, almost all B8R-specific memory CD8 T cells (CD44hi B8R+) from C57BL/6 mice infected with VV express NKG2D (Fig. 1D). Following infection with VV, the frequency of CD44hi effector or memory CD8 T cells expressing NKG2D was increased by >100-fold in the effector phase leading to a 10-fold expansion of the NKG2D+ subset in the memory phase (Fig. 1E, Supplemental Fig. 3). This was not specific to VV because at 55 d postinfection by Flu or bacteria L. monocytogenes, there was also a strong amplification of the NKG2D+ CD44hi CD8 T population in the blood (Fig. 1E). In contrast, the number of memory-phenotype NKG2D cells remained stable when comparing the naive and the memory phase. Importantly, NKG2D cells contained a fraction (∼25%) of T effector memory CD8 T cells. The fraction of T effector memory cells among NKG2D+ memory cells decreased with time following activation to reach on average 40% of memory cells 6 wk postinfection (data not shown). These results could be extended to other mouse strains: in BALB/c mice and in the outbred mouse strain OF1 VV infection induced the generation of NKG2D+ memory CD8 T cells (Supplemental Fig. 4).

FIGURE 1.

NKG2D expression identifies polyclonal AI memory CD8 T cell populations. (A) CD45.2 OT-I or F5 TCR transgenic naive CD8 T cells were adoptively transferred in immunocompetent congenic mice. The next day, hosts received i.p. injections of the indicated γc cytokine Ab complex, as described in the 2Materials and Methods. Thirty days after transfer, the expression of NKG2D by OT-I CD8 T cells was assessed by flow cytometry. Black histogram: OT-I CD8 T cells, gray histogram: host’s CD8 T cells. The percentage of transferred CD45.2 cells expressing NKG2D is indicated (IL-2: 1 experiment, n = 3 mice; IL-4: 3 experiments, n = 10 mice). (B) NKG2D expression by memory-phenotype (CD44hi) spleen CD8 T cells from 6-wk naive C57BL/6 mice. Black histogram: CD44hi CD8 T cells, gray histogram: naive CD8 T cells. The percentage of NKG2D+ cells among CD44hi CD8 T cells is shown (two experiments, n = 12 mice). (C) NKG2D expression by VM CD8 T cells (B8R-specific CD44hi CD49 from spleen of nonimmunized naive C57BL/6 mice). Black histogram: VM CD8 T cells, gray histogram: naive CD8 T cells. The percentage of VM cells expressing NKG2D is indicated (two experiments, n = 10 mice). (D) B8R-specific memory CD8 T cells (B8R+) from VV-immunized C57BL/6 mice (>100 d postinfection) were assessed for their expression of NKG2D (black histogram). Naive CD8 T cells from same mice were used as control. The percentage of NKG2D+ cells among B8R-specific cells is shown (five experiments, n = 5–7 mice per experiment). (E) C57BL/6 mice were infected with VV (i.n.), Flu (i.n.) or L. monocytogenes (i.v.), and the number of total, NKG2D, and NKG2D+, CD44hi CD8 T cells was measured in the blood 55 d following pathogen infection. Graph shows mean expansion index (± SD) of the indicated CD44hi CD8 T cell populations. The dotted line represents an absence of expansion (index = 1). One representative experiment out of two (n = at least 6 mice per group and per experiment). *p < 0.05, **p < 0.01, Wilcoxon matched-pairs signed rank test. (F) NKG2D+ and NKG2D memory CD8 T cells were sorted from VV-infected C57BL/6 mice (55 d postinfection). Cells were stimulated for 6 h with VV-infected DC2.4 cells (DC + VV) or with PMA/ionomycin (PMA/iono) in the presence of GolgiStop. The percentage of IFN-γ+ cells was measured by intracellular cytokine staining. Graph shows the mean percentage (± SD) of IFN-γ+ cells among each cell population. One representative experiment out of three (n = 5 mice per experiment). **p < 0.01, Mann–Whitney U test. (G) NKG2D+ and NKG2D memory CD8 T cells were sorted from VV-infected CD45.1 C57BL/6 mice (50 d postinfection) and 105 cells were transferred in separate CD45.2 C57BL/6 congenic mice. The next day, hosts were infected with VV or L. monocytogenes or left uninfected. Graphs show the mean numbers (± SD) of CD45.1 donor cells recovered 7 d postinfection in the indicated organs. One representative experiment out of four (n = 6 mice per group and per experiment). **p < 0.01, Mann–Whitney U test. (H) Splenic naive, NKG2D+ and NKG2D memory CD8 T cells were cell-sorted from Flu-infected C57BL/6 mice (45 d postinfection) and 105 cells were transferred in separate host mice. The next day, hosts were infected with a lethal dose of Flu. Graph shows the percentage of survival observed among each group of mice. One representative experiment out of two (n = 10 mice per group and per experiment). *p < 0.05, log-rank (Mantel–Cox) test. ns, not significant.

FIGURE 1.

NKG2D expression identifies polyclonal AI memory CD8 T cell populations. (A) CD45.2 OT-I or F5 TCR transgenic naive CD8 T cells were adoptively transferred in immunocompetent congenic mice. The next day, hosts received i.p. injections of the indicated γc cytokine Ab complex, as described in the 2Materials and Methods. Thirty days after transfer, the expression of NKG2D by OT-I CD8 T cells was assessed by flow cytometry. Black histogram: OT-I CD8 T cells, gray histogram: host’s CD8 T cells. The percentage of transferred CD45.2 cells expressing NKG2D is indicated (IL-2: 1 experiment, n = 3 mice; IL-4: 3 experiments, n = 10 mice). (B) NKG2D expression by memory-phenotype (CD44hi) spleen CD8 T cells from 6-wk naive C57BL/6 mice. Black histogram: CD44hi CD8 T cells, gray histogram: naive CD8 T cells. The percentage of NKG2D+ cells among CD44hi CD8 T cells is shown (two experiments, n = 12 mice). (C) NKG2D expression by VM CD8 T cells (B8R-specific CD44hi CD49 from spleen of nonimmunized naive C57BL/6 mice). Black histogram: VM CD8 T cells, gray histogram: naive CD8 T cells. The percentage of VM cells expressing NKG2D is indicated (two experiments, n = 10 mice). (D) B8R-specific memory CD8 T cells (B8R+) from VV-immunized C57BL/6 mice (>100 d postinfection) were assessed for their expression of NKG2D (black histogram). Naive CD8 T cells from same mice were used as control. The percentage of NKG2D+ cells among B8R-specific cells is shown (five experiments, n = 5–7 mice per experiment). (E) C57BL/6 mice were infected with VV (i.n.), Flu (i.n.) or L. monocytogenes (i.v.), and the number of total, NKG2D, and NKG2D+, CD44hi CD8 T cells was measured in the blood 55 d following pathogen infection. Graph shows mean expansion index (± SD) of the indicated CD44hi CD8 T cell populations. The dotted line represents an absence of expansion (index = 1). One representative experiment out of two (n = at least 6 mice per group and per experiment). *p < 0.05, **p < 0.01, Wilcoxon matched-pairs signed rank test. (F) NKG2D+ and NKG2D memory CD8 T cells were sorted from VV-infected C57BL/6 mice (55 d postinfection). Cells were stimulated for 6 h with VV-infected DC2.4 cells (DC + VV) or with PMA/ionomycin (PMA/iono) in the presence of GolgiStop. The percentage of IFN-γ+ cells was measured by intracellular cytokine staining. Graph shows the mean percentage (± SD) of IFN-γ+ cells among each cell population. One representative experiment out of three (n = 5 mice per experiment). **p < 0.01, Mann–Whitney U test. (G) NKG2D+ and NKG2D memory CD8 T cells were sorted from VV-infected CD45.1 C57BL/6 mice (50 d postinfection) and 105 cells were transferred in separate CD45.2 C57BL/6 congenic mice. The next day, hosts were infected with VV or L. monocytogenes or left uninfected. Graphs show the mean numbers (± SD) of CD45.1 donor cells recovered 7 d postinfection in the indicated organs. One representative experiment out of four (n = 6 mice per group and per experiment). **p < 0.01, Mann–Whitney U test. (H) Splenic naive, NKG2D+ and NKG2D memory CD8 T cells were cell-sorted from Flu-infected C57BL/6 mice (45 d postinfection) and 105 cells were transferred in separate host mice. The next day, hosts were infected with a lethal dose of Flu. Graph shows the percentage of survival observed among each group of mice. One representative experiment out of two (n = 10 mice per group and per experiment). *p < 0.05, log-rank (Mantel–Cox) test. ns, not significant.

Close modal

To further validate NKG2D as a marker of AI memory T cells, we next analyzed NKG2D expression by VV-specific memory T cells recognizing other epitopes than B8R. Indeed, VV harbors at least 40 epitopes recognized by CD8 T cells (37). To extend this analysis to the whole viral epitope repertoire, memory CD8 T cells from VV-immunized mice were restimulated with VV-infected dendritic cells. More than half of NKG2D+ memory-phenotype CD8 T cells produced IFN-γ following VV restimulation, whereas only ∼5% of NKG2D memory-phenotype CD8 T cells did (Fig. 1F). This lack of IFN-γ production was not due to a functional defect of NKG2D memory-phenotype CD8 T cells, as restimulation with PMA and ionomycin led to IFN-γ production. As some epitopes of VV might be expressed in a delayed fashion and might not be presented in the time frame used for in vitro restimulation, we also performed in vivo rechallenge of memory CD8 T cells (Fig. 1G). Although NKG2D and NKG2D+ memory CD8 T cells from VV-infected mice equally grafted in the spleen of host mice, only NKG2D+ memory-phenotype CD8 T cells had strongly proliferated 7 d after rechallenge with VV, as revealed by the number of donor cells recovered in the spleen and lung of host mice (Fig. 1G). In contrast, the number of donor NKG2D memory-phenotype CD8 T cells remained close to the number observed in unimmunized host mice, revealing marginal expansion following VV rechallenge. NKG2D+ memory CD8 T cell proliferation was strictly dependent on Ag recognition, as infection with the heterologous pathogen L. monocytogenes did not lead to their expansion (Fig. 1G). Finally, we compared the capacity of NKG2D+ and NKG2D memory-phenotype CD8 T cells to protect naive mice against a lethal dose of virus. To do so, naive CD8 T cells as well as NKG2D and NKG2D+ memory CD8 T cells from Flu-immune mice were transferred in naive hosts that were infected with a lethal dose of Flu. NKG2D+ memory CD8 T cells induced a significant protection of host mice as more than 40% of them survived the infection (Fig. 1H). This is in contrast to naive cells and NKG2D memory CD8 T cells that conferred no protection. Altogether, these results indicate that NKG2D expression identifies AI memory CD8 T cells in polyclonal settings, allowing them to be distinguished from IN memory CD8 T cells that do not express this marker.

Next, taking advantage of NKG2D as a marker discriminating AI and IN memory CD8 T cells, we compared these two populations in terms of TCR repertoire and effector functions. A multiplex PCR identifying β-chain locus V(D)J rearrangements showed that, as expected, AI memory CD8 T cells have a less diverse TCR repertoire than naive CD8 T cells, reflecting Ag selection (Fig. 2A). In contrast, IN memory CD8 T cells have a TCR repertoire that is as diverse as that of naive CD8 T cells. A principal component (PC) analysis indicated that the TCR repertoire was more similar within the NKG2D+ subset of different mice than within the CD8 cells (i.e., the naive, NKG2D+, and NKG2D cells of a given mouse) (Fig. 2A). As IN cells have a diverse repertoire that seems to differ in its composition from naive CD8 T cells, we tested if IN cells could participate in a primary immune response against a pathogen (i.e., whether they can mount responses against unknown foreign epitopes) and compared their responsiveness to naive cells. Equal numbers of CD45.1 naive and CD45.2 IN memory CD8 T cells were sorted from naive mice and cotransferred to congenic CD45.1/CD45.2 hosts that were infected the next day with VV or L. monocytogenes. Seven days postinfection, the contribution of transferred cell populations to the primary response was determined in the spleen (Fig. 2B). Both naive and IN memory cells expanded during the primary immune response and became NKG2D+. We also evaluated the relative contribution of the two grafted cell populations to the NKG2D+ CD44hi effector cells response. During VV infection, naive and IN memory CD8 T cells generated almost equal numbers of NKG2D+ CD44hi effector cells. In contrast, during L. monocytogenes infection, IN memory CD8 T cells generated more NKG2D+ CD44hi effector cells compared with naive cells (Fig. 2B). In conclusion, IN cells have a more diversified repertoire than AI cells and can contribute to a primary T cell response.

FIGURE 2.

Characterization of IN memory CD8 T cells. (A) Naive, IN and AI memory CD8 T cells were sorted from VV-infected C57BL/6 mice (50 d postinfection), and multiplex PCR was performed on genomic DNA to detect VβJβ rearrangements among each cell population. Left graph shows the mean percentage (± SD) of TCR repertoire diversity, calculated as described in 2Materials and Methods. *p < 0.05, **p < 0.01, Mann–Whitney U test (n = 5 mice). Right: PC analysis based on the presence or absence of each possible VβJβ rearrangement among each cell population. This representation shows the distribution of CD8 T cell populations regarding to the composition of their VβJβ TCR repertoire. Each form of symbol represents a mouse and lines connect cell populations sorted from the same mouse. (B) Naive and IN memory CD8 T cells were sorted from naive CD45.1/CD45.2 and CD45.2 C57BL/6 mice respectively and were cotransferred at a 1:1 ratio in congenic CD45.1 C57BL/6 mice. Host mice were infected with VV or L. monocytogenes or left uninfected (Ø). Left graph shows the mean number (± SD) of CD44hi NKG2D+ CD8 T cells in the spleen generated from transferred cell populations 7 d postinfection. Right graph shows the contribution of each transferred CD8 T cell population to the total CD44hi NKG2D+ CD8 T cells generated from transferred cells (mean percentage ± SD). One representative experiment out of two (n = 5 mice per group and per experiment). *p < 0.05, **p < 0.01, Mann–Whitney U test. (C) Equal number of naive, IN, and AI memory CD8 T cells were sorted from VV-infected mice (>100 d postinfection) and were restimulated for 12 h with anti-CD3 plus anti-CD28 Abs in the presence of IL-2. Radar plots show the mean amount of each cytokine produced, measured by multiplex. One experiment (n = 5 mice). (D) Graph shows the mean amount (± SD) of memory-associated poised cytokines produced by CD8 T cell populations, measured by ELISA. One representative experiment out of three (n = 5 mice per experiment). **p < 0.01, Mann–Whitney U test. (E) Production of IFN-γ by naive, IN, and AI memory CD8 T cells from VV-infected mice (50 d postinfection) was measured by intracellular cytokine staining following 5 h of restimulation with anti-CD3 plus anti-CD28 Abs or with a mixture of IL-12/IL-18/IL-2. Graph shows the mean percentage (± SD) of IFN-γ+ cells among each CD8 T cell population. A pool of two representative experiments is shown (n = 4 mice in total). *p < 0.05, Mann–Whitney U test. ND, not detected; ns, not significant.

FIGURE 2.

Characterization of IN memory CD8 T cells. (A) Naive, IN and AI memory CD8 T cells were sorted from VV-infected C57BL/6 mice (50 d postinfection), and multiplex PCR was performed on genomic DNA to detect VβJβ rearrangements among each cell population. Left graph shows the mean percentage (± SD) of TCR repertoire diversity, calculated as described in 2Materials and Methods. *p < 0.05, **p < 0.01, Mann–Whitney U test (n = 5 mice). Right: PC analysis based on the presence or absence of each possible VβJβ rearrangement among each cell population. This representation shows the distribution of CD8 T cell populations regarding to the composition of their VβJβ TCR repertoire. Each form of symbol represents a mouse and lines connect cell populations sorted from the same mouse. (B) Naive and IN memory CD8 T cells were sorted from naive CD45.1/CD45.2 and CD45.2 C57BL/6 mice respectively and were cotransferred at a 1:1 ratio in congenic CD45.1 C57BL/6 mice. Host mice were infected with VV or L. monocytogenes or left uninfected (Ø). Left graph shows the mean number (± SD) of CD44hi NKG2D+ CD8 T cells in the spleen generated from transferred cell populations 7 d postinfection. Right graph shows the contribution of each transferred CD8 T cell population to the total CD44hi NKG2D+ CD8 T cells generated from transferred cells (mean percentage ± SD). One representative experiment out of two (n = 5 mice per group and per experiment). *p < 0.05, **p < 0.01, Mann–Whitney U test. (C) Equal number of naive, IN, and AI memory CD8 T cells were sorted from VV-infected mice (>100 d postinfection) and were restimulated for 12 h with anti-CD3 plus anti-CD28 Abs in the presence of IL-2. Radar plots show the mean amount of each cytokine produced, measured by multiplex. One experiment (n = 5 mice). (D) Graph shows the mean amount (± SD) of memory-associated poised cytokines produced by CD8 T cell populations, measured by ELISA. One representative experiment out of three (n = 5 mice per experiment). **p < 0.01, Mann–Whitney U test. (E) Production of IFN-γ by naive, IN, and AI memory CD8 T cells from VV-infected mice (50 d postinfection) was measured by intracellular cytokine staining following 5 h of restimulation with anti-CD3 plus anti-CD28 Abs or with a mixture of IL-12/IL-18/IL-2. Graph shows the mean percentage (± SD) of IFN-γ+ cells among each CD8 T cell population. A pool of two representative experiments is shown (n = 4 mice in total). *p < 0.05, Mann–Whitney U test. ND, not detected; ns, not significant.

Close modal

One key property of memory CD8 T cells is the rapid and increased production of cytokines and chemokines upon TCR stimulation. We therefore compared the capacity of polyclonal IN and AI memory and naive CD8 T cells to produce cytokines and chemokines in response to TCR stimulation. Early after TCR triggering, both IN and AI cells produced a broader array of cytokines/chemokines than naive cells. These factors were also secreted in larger amounts by memory T cells of both subsets (Fig. 2C). Importantly, the same pattern of cytokines/chemokines was produced by IN and AI memory cells, although AI cells produced at least a 10-fold higher quantity of the poised memory cytokine/chemokines CCL5, CCL1, and IFN-γ than IN cells (Fig. 2D). AI memory cells can also produce IFN-γ upon stimulation by innate signals such as IL-12 and IL-18, a property known as the IN function of memory CD8 T cells (38). A fraction of IN memory cells produced IFN-γ upon IL-12/IL-18 stimulation, albeit at a lower frequency than AI memory CD8 T cells (Fig. 2E). Taken together, these results indicate that although AI and IN memory CD8 T cells produced a similar pattern of cytokines following stimulation, AI produce higher amounts of cytokines than IN memory T cells.

To further characterize the two subsets defined by NKG2D expression, we compared their transcriptome using microarrays. F5 TCR transgenic CD8 T cells were transferred in naive mice before infection with VV-NP68 to establish an internal control of AI memory CD8 T cells. Eighty days postinfection, F5 memory cells as well as host’s IN and AI memory CD8 T cell populations were sorted, and their transcriptome was analyzed. Naive CD8 T cells were sorted from naive F5 and C57BL/6 mice. We first performed a PC analysis that shows that 70% of the variability of the samples is explained by the first two PC. Samples were aligned along PC1 according to their differentiation stage, whereas PC2 highlighted a difference between monoclonal F5 TCR transgenic T cells and polyclonal CD8 T cells (Fig. 3A). F5 and polyclonal CD8 T cells differed by the expression of few genes, among which those coding for the TCR, reflecting the monoclonality of the F5 CD8 population (Supplemental Fig. 5A), and one set of genes encoding the inhibitory NK cell receptors (Klra) were expressed by polyclonal AI memory cells but not F5 memory cells. Accordingly, these Ly49 receptors are exclusively expressed by a small fraction of polyclonal memory CD8 T cells in contrast to TCR transgenic F5 memory cells (Supplemental Fig. 5B). This property is shared by both IN and AI polyclonal memory cells (Supplemental Fig. 5C). Importantly, all genes differentially expressed by AI memory CD8 T cells compared with naive cells were also differentially expressed by F5 CD8 T cells when comparing memory to naive cells, confirming their AI nature (data not shown). We then compared AI and IN polyclonal memory populations. IN memory cells were positioned closer to AI memory cells than to naive CD8 T cells on the PC1 axis, confirming their memory differentiation (Fig. 3A). Indeed, AI and IN memory CD8 T cell transcriptomes differ in the expression levels of several genes encoding transcription factors, effector molecules, NK cell receptors, chemokine receptors, and integrins (Fig. 3B). Compared with IN cells, AI memory CD8 T cells express higher levels of genes encoding effector molecules involved in the killing of target cells through cytotoxicity. Accordingly, this memory cell population displays a higher capacity to mediate killing, in anti-CD3 redirected cytotoxicity assay (Supplemental Fig. 6). Importantly, AI memory CD8 T cells express higher levels than IN cells of transcription factors that promote the full differentiation of memory CD8 T cells, such as Tbet, ID2, Zeb2, and Blimp-1 (Fig. 3C, 3D). In contrast, IN memory CD8 T cells show increased levels of transcription factors that promote a less differentiated state of memory CD8 T cells, such as Eomes and ID3. The pattern of transcription factors expressed by AI memory cells is thus in accordance with their more differentiated state compared with IN.

FIGURE 3.

Transcriptome analysis of AI and IN memory CD8 T cells. Naive CD45.1 F5 TCR transgenic CD8 T cells were adoptively transferred in congenic host mice that were subsequently immunized with VV-NP68. Eighty days postinfection, F5 (TCR transgenic) memory cells and polyclonal NKG2D and NKG2D+ memory CD8 T cells from the host were sorted from five independent groups of eight mice. As a control, naive F5 and polyclonal CD8 T cells were sorted from five independent groups of three naive F5 and C57BL/6 mice, respectively. The transcriptome of these cell populations was compared by microarrays. (A) PC analysis was performed on whole microarray data. Left graph shows the distribution of samples according to PC1 and PC2. Right graph shows the percentage of variance explained by successive PC. (B) The main genes differently expressed between polyclonal IN and AI memory CD8 T cells are listed. Fold changes and p values are indicated for each gene. The family to which each group of genes belongs is also indicated. (C) Naive, IN, and AI memory CD8 T cells were sorted from VV-infected mice (50 d postinfection). The expression level of several transcription factors by each cell population was assessed by quantitative PCR. Graph shows the mean fold increase (± SD) compared with naive CD8 T cells. One representative experiment out of two (n = 5 mice per experiment). (D) The expression of Tbet and Eomes by naive, IN, and AI memory CD8 T cells from VV-infected mice (50 d postinfection) was assessed by flow cytometry. Graphs show mean of mean fluorescence intensity (± SD) of each transcription factor. One representative experiment out of two (n = 4 mice per experiment). *p < 0.05, Mann–Whitney U test.

FIGURE 3.

Transcriptome analysis of AI and IN memory CD8 T cells. Naive CD45.1 F5 TCR transgenic CD8 T cells were adoptively transferred in congenic host mice that were subsequently immunized with VV-NP68. Eighty days postinfection, F5 (TCR transgenic) memory cells and polyclonal NKG2D and NKG2D+ memory CD8 T cells from the host were sorted from five independent groups of eight mice. As a control, naive F5 and polyclonal CD8 T cells were sorted from five independent groups of three naive F5 and C57BL/6 mice, respectively. The transcriptome of these cell populations was compared by microarrays. (A) PC analysis was performed on whole microarray data. Left graph shows the distribution of samples according to PC1 and PC2. Right graph shows the percentage of variance explained by successive PC. (B) The main genes differently expressed between polyclonal IN and AI memory CD8 T cells are listed. Fold changes and p values are indicated for each gene. The family to which each group of genes belongs is also indicated. (C) Naive, IN, and AI memory CD8 T cells were sorted from VV-infected mice (50 d postinfection). The expression level of several transcription factors by each cell population was assessed by quantitative PCR. Graph shows the mean fold increase (± SD) compared with naive CD8 T cells. One representative experiment out of two (n = 5 mice per experiment). (D) The expression of Tbet and Eomes by naive, IN, and AI memory CD8 T cells from VV-infected mice (50 d postinfection) was assessed by flow cytometry. Graphs show mean of mean fluorescence intensity (± SD) of each transcription factor. One representative experiment out of two (n = 4 mice per experiment). *p < 0.05, Mann–Whitney U test.

Close modal

One important characteristic of memory cells is their capacity to circulate and migrate to inflamed tissues (3941). As IN cells express fewer memory-specific chemokine receptors and integrins than AI cells, we compared their ability to traffic to the lung upon inflammation or infection. Inflammation was first induced in mouse lungs by the TLR3 agonist poly(I:C) that induces the production of type-I IFN and its downstream chemokines (CXCL9, 10, and 11) (42). Memory CD8 T cells containing similar numbers of IN and AI populations were purified from VV-infected mice and transferred into host mice that then received i.n. injection of poly(I:C). The recruitment of AI and IN memory cells in different organs was assessed 2 d after poly(I:C) injection. The same numbers of donor AI and IN memory CD8 T cells were found in the spleen and in the blood of recipient mice injected with PBS or poly(I:C) (Supplemental Fig. 7), as reflected by a cell number ratio close to 1 (Fig. 4A). As expected the recruitment of donor memory CD8 T cells within the lung parenchyma and airways was induced following poly(I:C) injection (Supplemental Fig. 7). However, AI memory CD8 T cells were preferentially recruited compared with IN cells (Fig. 4A). Following lung infection by a pathogen, spleen memory cells are recruited to the lung in an Ag-independent fashion (43, 44). To analyze the capacity of AI and IN memory CD8 T cells to be attracted to the infected lung, VV-specific memory CD8 T cells containing NKG2D and NKG2D+ populations were isolated from VV-immune mice and transferred into host mice that had been immunized 2 d before with Flu virus, and the recruitment was assessed 2 d later. Again, AI memory CD8 T cells were preferentially recruited to the lung during Flu infection, accumulating within the parenchyma (Fig. 4B). In conclusion, our results show that upon inflammation or infection, AI memory CD8 T cells enter the lung parenchyma more efficiently than IN memory CD8 T cells.

FIGURE 4.

IN cells have a reduced capacity to access inflamed peripheral tissues compared with AI memory CD8 T cells. (A) CD45.2 memory CD8 T cells were purified from VV-infected mice (i.p. immunization) and transferred into congenic CD45.1 mice. The next day, host mice received i.n. administration of poly(I:C) or PBS. Two days later, the numbers of donor IN and AI memory CD8 T cells were measured in various organs. Graph shows the mean ratios (± SD) between donor IN and AI memory CD8 T cells in spleen, blood, and lung. Lung intravascular staining was performed to differentiate cells in the vasculature and in the parenchyma. One representative experiment out of two (n = 5 mice per group per experiment). (B) CD45.2 memory CD8 T cells were purified from VV-infected mice (i.p. immunization) and transferred into congenic CD45.1 mice immunized with Flu virus (i.n. immunization) 2 d earlier. Graphs show mean numbers (± SD) of donor IN and AI memory CD8 T cells in the spleen, the blood, and the lung (total or in the parenchyma) 2 d later. **p < 0.01, Mann–Whitney U test.

FIGURE 4.

IN cells have a reduced capacity to access inflamed peripheral tissues compared with AI memory CD8 T cells. (A) CD45.2 memory CD8 T cells were purified from VV-infected mice (i.p. immunization) and transferred into congenic CD45.1 mice. The next day, host mice received i.n. administration of poly(I:C) or PBS. Two days later, the numbers of donor IN and AI memory CD8 T cells were measured in various organs. Graph shows the mean ratios (± SD) between donor IN and AI memory CD8 T cells in spleen, blood, and lung. Lung intravascular staining was performed to differentiate cells in the vasculature and in the parenchyma. One representative experiment out of two (n = 5 mice per group per experiment). (B) CD45.2 memory CD8 T cells were purified from VV-infected mice (i.p. immunization) and transferred into congenic CD45.1 mice immunized with Flu virus (i.n. immunization) 2 d earlier. Graphs show mean numbers (± SD) of donor IN and AI memory CD8 T cells in the spleen, the blood, and the lung (total or in the parenchyma) 2 d later. **p < 0.01, Mann–Whitney U test.

Close modal

Memory CD8 T cell trafficking to the lung parenchyma and airways has been reported to be dependent on ITGA1-4 integrins and chemokine receptors CXCR3 (39, 41, 45). Using CXCR3 knockout (KO) mice, we confirmed that CXCR3 expression is required for AI memory cell recruitment to the lung parenchyma (Fig. 5A). However, the differential recruitment of AI and IN memory subsets did not involve CXCR3, as both memory cell types expressed a similar level of CXCR3 and migrated strongly toward CXCL10 in a transwell assay (Fig. 5B). By contrast, AI cells differ from IN cells by the expression level of several integrin mRNA; namely, Itga1, Itga4, and Itgb1 encoding for the α- and β-chains of VLA1 (CD49a, CD29) and VLA4 (CD49d, CD29), respectively (Fig. 3B) (26, 46), a difference confirmed at the protein level by flow cytometry analysis (Fig. 5C). We thus tested if Abs directed against CD49a and CD49d could alter the recruitment of AI cells to the infected lung. Ab treatment did not affect AI cell numbers in spleen and blood (Supplemental Fig. 8). In contrast, it significantly inhibited AI cell recruitment to the lung parenchyma (Fig. 5D). Altogether, these results indicate that following lung infection, AI memory cells are the main memory subset recruited to the lung parenchyma, a process that is dependent on CXCR3 and CD49a/CD49d integrin expression.

FIGURE 5.

Role of CXCR3 and CD49a/CD49d for entry of AI memory CD8 T cells into inflamed lung. (A) Memory CD8 T cells purified from VV-infected mice, C57BL/6 mice, or CXCR3 KO mice (i.p. immunization) were transferred into congenic hosts. Thirty-five days later, host mice received i.n. administration of Flu virus. Two days later, the number of donor IN and AI memory CD8 T cells was measured in lung parenchyma. One representative experiment out of two. *p < 0.05, Mann–Whitney U test. (B) Histogram shows the expression of CXCR3 by naive, IN, and AI memory CD8 T cells from VV-infected mice (50 d postinfection). Gray histogram: control isotype. Graph shows the migration of naive, IN, and AI memory CD8 T cells from VV-infected mice toward CXCL10 in a transwell assay (mean ± SD). One representative experiment out of three (n = 5 mice per experiment). *p < 0.05, **p < 0.01, Mann–Whitney U test. (C) The expression of CD29, CD49a, and CD49d by naive, IN, and AI memory CD8 T cells from VV-infected mice (50 d postinfection) was assessed by flow cytometry. Gray histogram: control isotype. One representative experiment out of three (n = 2 mice per experiment). (D) Memory CD8 T cells were purified from VV-infected mice (i.n. immunization), incubated or not with 1 μg/ml anti-CD49a and anti-CD49d Abs, and then transferred into congenic mice that had been immunized i.n. 2 d before with Flu virus. The next day, host mice received i.p. administration of 250 μg anti-CD49a plus anti-CD49d Abs or PBS. Two days later, the number of donor IN and AI memory CD8 T cells was measured in lung parenchyma. Graph shows mean numbers (± SD) of donor IN and AI memory CD8 T cells. A pool of two representative experiments is shown (n = 13 mice in total). **p < 0.01, Mann–Whitney U test.

FIGURE 5.

Role of CXCR3 and CD49a/CD49d for entry of AI memory CD8 T cells into inflamed lung. (A) Memory CD8 T cells purified from VV-infected mice, C57BL/6 mice, or CXCR3 KO mice (i.p. immunization) were transferred into congenic hosts. Thirty-five days later, host mice received i.n. administration of Flu virus. Two days later, the number of donor IN and AI memory CD8 T cells was measured in lung parenchyma. One representative experiment out of two. *p < 0.05, Mann–Whitney U test. (B) Histogram shows the expression of CXCR3 by naive, IN, and AI memory CD8 T cells from VV-infected mice (50 d postinfection). Gray histogram: control isotype. Graph shows the migration of naive, IN, and AI memory CD8 T cells from VV-infected mice toward CXCL10 in a transwell assay (mean ± SD). One representative experiment out of three (n = 5 mice per experiment). *p < 0.05, **p < 0.01, Mann–Whitney U test. (C) The expression of CD29, CD49a, and CD49d by naive, IN, and AI memory CD8 T cells from VV-infected mice (50 d postinfection) was assessed by flow cytometry. Gray histogram: control isotype. One representative experiment out of three (n = 2 mice per experiment). (D) Memory CD8 T cells were purified from VV-infected mice (i.n. immunization), incubated or not with 1 μg/ml anti-CD49a and anti-CD49d Abs, and then transferred into congenic mice that had been immunized i.n. 2 d before with Flu virus. The next day, host mice received i.p. administration of 250 μg anti-CD49a plus anti-CD49d Abs or PBS. Two days later, the number of donor IN and AI memory CD8 T cells was measured in lung parenchyma. Graph shows mean numbers (± SD) of donor IN and AI memory CD8 T cells. A pool of two representative experiments is shown (n = 13 mice in total). **p < 0.01, Mann–Whitney U test.

Close modal

In this study, we demonstrated that NKG2D surface expression is restricted to AI memory CD8 T cells, allowing their discrimination from IN memory CD8 T cells generated under physiological conditions. We showed that this dichotomy is conserved in different mouse strains and in response to infection by different pathogens. Expression of NKG2D by VM CD8 T cells has recently been reported at the mRNA level (46). We found that sorted NKG2D memory-phenotype IN cells containing the VM cells expressed higher levels of mRNA encoding NKG2D compared with naive cells (fold change = 2, Supplemental Fig. 9). However, we found that the small fraction of B8R+ VM cells (0.02%) within CD44hi CD49d CD8 T cells found in naive mice are mainly NKG2D at the protein level (Fig. 1C). Importantly, NKG2D+ AI memory CD8 T cells expressed much higher levels of NKG2D mRNA than IN memory cells in quiescent cells, which could explain the difference observed in NKG2D protein levels. Thus, the surface expression of NKG2D protein is restricted to AI memory CD8 T cells.

Taking advantage of NKG2D expression, we compared IN memory CD8 T cells to pathogen-induced memory CD8 T cells. We performed a transcriptome comparison of IN and AI memory CD8 T cell populations. Our results clearly demonstrated that VV-induced memory cells, whether polyclonal (NKG2D+ CD8 T cells) or monoclonal (F5 TCR transgenic CD8 T cells), share the same transcriptome. IN cells have also acquired a genetic program typical of memory; nevertheless, these cells are less differentiated compared with AI memory CD8 T cells. In agreement with a previous study (47), IN cells did not express a specific gene expression pattern that could indicate an independent differentiation pathway. This suggests that IN memory CD8 T cells represent an intermediate stage of differentiation between naive and AI memory CD8 T cells rather than a distinct CD8 T cell lineage. This is also supported by their cytokine secretion pattern in response to TCR stimulation, which is similar to that of AI memory CD8 T cells. Differentiation of memory cells is regulated by different pairs of transcription factors, such as Tbet and Eomes, Blimp1/Bcl6, or ID2/ID3 (48). Interestingly, genes encoding for these transcription factors are differentially expressed between IN and AI memory CD8 T cells. Indeed, AI memory CD8 T cells express higher levels of genes encoding for transcription factors that promote memory CD8 T cell full differentiation (Tbx21, Id2, Prdm1, Zeb2). In contrast, IN memory CD8 T cells express higher levels of genes encoding transcription factors that favor a less differentiated state (Id3, Eomes). Thus, the expression pattern of transcription factors observed in IN and AI memory CD8 T cell populations fits with the observed differentiation state. In agreement with their transcription factors expression pattern, AI memory CD8 T cells express higher levels of genes encoding for effector molecules, such as granzymes, perforin, and Fas ligand, confirming that these cells are more differentiated and are able to kill a potential target more rapidly.

One major difference between naive and memory cells is the capacity of memory cells to access and reside within tissues parenchyma. Following a pulmonary infection, pathogen-specific CD8 effector cells enter the lung parenchyma, and under specific signaling, some of them differentiate in TRM that ensure a robust response if a secondary infection occurs (3, 4). In the lung, in contrast to other tissues, the population of TRM wanes over time (5, 6). Thus, long-term protection of the lung relies on the recruitment of secondary memory cells stored in lymphoid tissues. Indeed, following infection of the lung, inflammation rapidly induces the recruitment of memory cells independently of their Ag specificity (43, 44, 49), although pathogen-specific spleen memory T cells also rapidly gain access to the tissue (2).

Previous studies have demonstrated that the CD49d integrin is differentially expressed between B8R-specific IN and AI memory CD8 T cells (22, 26). Transcriptome comparison of IN and AI cells revealed here that other integrin chains are also differentially expressed by AI and IN memory cells. This was confirmed at the protein level: AI memory CD8 T cells expressed higher levels of several integrins (CD29, CD49a, CD49d) compared with naive or IN memory CD8 T cells. These integrins play a key role in immune cell migration, allowing them to exit blood vessels and access peripheral tissues (45). Indeed, ITGA1/B2 (CD29, CD49a) and ITGA4/B2 (CD29, CD49d) play an essential role in the extravasation of CD8 T cells in the lung or the brain, respectively (50, 51). Accordingly, we found that upon lung inflammation, AI memory CD8 T cells were preferentially recruited within the lung parenchyma, and we demonstrated that this process was dependent on integrins. In contrast, IN memory CD8 T cells remained within the lung vasculature. Memory CD8 T cell trafficking to the lung parenchyma and airways has been reported to be dependent of the chemokine receptor CXCR3 (41). In agreement, CXCR3 KO AI memory cells were not recruited to the lung. Importantly, we did not observe any differential CXCR3 expression or CXCL10-induced migration between IN and AI memory CD8 T cells. This indicates that CXCR3 expression is necessary but not sufficient to access the inflamed lung parenchyma.

IN memory CD8 T cells have a TCR repertoire as diverse as that of naive CD8 T cells, although their repertoire does not completely overlap. This diversified TCR repertoire could allow IN memory CD8 T cells to contribute to multiple immune responses. In line with this, we showed that IN memory CD8 T cells participated in primary immune responses against two pathogens, namely VV and L. monocytogenes. Participation of IN memory CD8 T cells in primary immune responses against infectious pathogens could significantly increase the efficiency of these responses. Indeed, we showed that physiologically generated IN cells produced a lot more cytokines than naive cells when stimulated through their TCR. In line with this, Lee et al. (27) demonstrated that OVA-specific IN memory CD8 T cells cleared L. monocytogenes–OVA infection more efficiently than naive CD8 T cell do. Similarly, results obtained in mice that are deficient in IL-4–induced IN cells indicate a decreased capacity to control a primary lymphocytic choriomeningitis virus infection in the absence of these cells (52). The protection conferred by IN cells could be direct, as a result of their enhanced effector functions, or indirect, through the help to naive CD8 T cells (53). However, our results show that they are not recruited to the inflamed lung parenchyma because of the lack of ITGA1/4 integrins expression. This is in line with the demonstration that VM CD8 T cells do not confer protection against L. monocytogenes upon gut infection (46).

The decreased capacity of IN memory cells to access nonlymphoid tissues in response to inflammatory chemokines could be essential for the prevention of autoimmunity. Because of its generation process, the population of IN memory CD8 T cells might be preferentially generated from naive cells with increased sensitivity for self-antigens. Indeed, CD5hi naive CD8 T cells, which have an increased sensitivity to self-antigens, are more prone to undergo LIP compared with CD5lo naive cells (54, 55). Moreover, CD5hi naive cells are more predisposed to become VM CD8 T cells compared with CD5lo naive cells. Of note, increased lymphopenia drives the development of autoaggressive T cells in NOD mice. This mechanism accounts partly for the development of diabetes in this mouse model (56). The exclusion from peripheral tissues of IN memory CD8 T cells that show increased reactivity compared with naive cells might thus be important to avoid autoimmunity.

In conclusion, NKG2D is a novel marker of AI memory cells. Moreover, although AI and IN memory CD8 T cells are similar from a phenotypic, transcriptomic, and functional point of view, they differ in their capacity to be recruited to inflamed lung parenchyma in an integrin-dependent fashion.

We thank Dr. Wencker for critical reading of the manuscript. We also acknowledge the contribution of SFR BioSciences (UMS3444-CNRS/US8-INSERM, École Normale Supérieure de Lyon, Université de Lyon) facilities and their staff, especially T. Andrieu and S. Dussurgey (AniRA-Cytométrie) and J.L. Thoumas, C. Angleraux and J.F. Henry (AniRA-PBES).

This work was supported by INSERM, CNRS, Université de Lyon, École Normale Supérieure de Lyon, Région Rhône-Alpes, and Agence Nationale de la Recherche (Grant 12-RPIB-0011). M.G. was a Région Rhône-Alpes Ph.D. fellow.

The microarray data presented in this article have been submitted to the Gene Expression Omnibus (http://www.ncbi.nlm.nih.gov/geo/) under accession number GSE111137.

The online version of this article contains supplemental material.

Abbreviations used in this article:

     
  • AI

    Ag-induced

  •  
  • Flu

    recombinant influenza

  •  
  • i.n.

    intranasal

  •  
  • IN

    innate

  •  
  • KO

    knockout

  •  
  • LIP

    lymphopenia-induced proliferation

  •  
  • PC

    principal component

  •  
  • SPF

    specific pathogen-free

  •  
  • TRM

    tissue-resident memory cell

  •  
  • VM

    virtual memory

  •  
  • VV

    vaccinia virus.

1
Pihlgren
,
M.
,
P. M.
Dubois
,
M.
Tomkowiak
,
T.
Sjögren
,
J.
Marvel
.
1996
.
Resting memory CD8+ T cells are hyperreactive to antigenic challenge in vitro.
J. Exp. Med.
184
:
2141
2151
.
2
Brinza
,
L.
,
S.
Djebali
,
M.
Tomkowiak
,
J.
Mafille
,
C.
Loiseau
,
P.-E.
Jouve
,
S.
de Bernard
,
L.
Buffat
,
B.
Lina
,
M.
Ottmann
, et al
.
2016
.
Immune signatures of protective spleen memory CD8 T cells.
Sci. Rep.
6
:
37651
.
3
Schenkel
,
J. M.
,
K. A.
Fraser
,
V.
Vezys
,
D.
Masopust
.
2013
.
Sensing and alarm function of resident memory CD8+ T cells. [Published erratum appears in 2013 Nat. Immunol. 14: 876.]
Nat. Immunol.
14
:
509
513
.
4
Mueller
,
S. N.
,
T.
Gebhardt
,
F. R.
Carbone
,
W. R.
Heath
.
2013
.
Memory T cell subsets, migration patterns, and tissue residence.
Annu. Rev. Immunol.
31
:
137
161
.
5
Wu
,
T.
,
Y.
Hu
,
Y.-T.
Lee
,
K. R.
Bouchard
,
A.
Benechet
,
K.
Khanna
,
L. S.
Cauley
.
2014
.
Lung-resident memory CD8 T cells (TRM) are indispensable for optimal cross-protection against pulmonary virus infection.
J. Leukoc. Biol.
95
:
215
224
6
Slütter
,
B.
,
N.
Van Braeckel-Budimir
,
G.
Abboud
,
S. M.
Varga
,
S.
Salek-Ardakani
,
J. T.
Harty
.
2017
.
Dynamics of influenza-induced lung-resident memory T cells underlie waning heterosubtypic immunity.
Sci Immunol.
2
:
eaag2031
.
7
Jameson
,
S. C.
,
Y. J.
Lee
,
K. A.
Hogquist
.
2015
.
Innate memory T cells.
Adv. Immunol.
126
:
173
213
.
8
Ernst
,
B.
,
D. S.
Lee
,
J. M.
Chang
,
J.
Sprent
,
C. D.
Surh
.
1999
.
The peptide ligands mediating positive selection in the thymus control T cell survival and homeostatic proliferation in the periphery.
Immunity
11
:
173
181
.
9
Goldrath
,
A. W.
,
M. J.
Bevan
.
1999
.
Low-affinity ligands for the TCR drive proliferation of mature CD8+ T cells in lymphopenic hosts.
Immunity
11
:
183
190
.
10
Goldrath
,
A. W.
,
L. Y.
Bogatzki
,
M. J.
Bevan
.
2000
.
Naive T cells transiently acquire a memory-like phenotype during homeostasis-driven proliferation.
J. Exp. Med.
192
:
557
564
.
11
Kieper
,
W. C.
,
S. C.
Jameson
.
1999
.
Homeostatic expansion and phenotypic conversion of naïve T cells in response to self peptide/MHC ligands.
Proc. Natl. Acad. Sci. USA
96
:
13306
13311
.
12
Murali-Krishna
,
K.
,
R.
Ahmed
.
2000
.
Cutting edge: naive T cells masquerading as memory cells.
J. Immunol.
165
:
1733
1737
.
13
Goldrath
,
A. W.
,
P. V.
Sivakumar
,
M.
Glaccum
,
M. K.
Kennedy
,
M. J.
Bevan
,
C.
Benoist
,
D.
Mathis
,
E. A.
Butz
.
2002
.
Cytokine requirements for acute and Basal homeostatic proliferation of naive and memory CD8+ T cells.
J. Exp. Med.
195
:
1515
1522
.
14
Jameson
,
S. C.
2002
.
Maintaining the norm: T-cell homeostasis.
Nat. Rev. Immunol.
2
:
547
556
.
15
Sprent
,
J.
,
C. D.
Surh
.
2011
.
Normal T cell homeostasis: the conversion of naive cells into memory-phenotype cells.
Nat. Immunol.
12
:
478
484
.
16
Kamimura
,
D.
,
M. J.
Bevan
.
2007
.
Naive CD8+ T cells differentiate into protective memory-like cells after IL-2 anti IL-2 complex treatment in vivo.
J. Exp. Med.
204
:
1803
1812
.
17
Lee
,
Y. J.
,
S. C.
Jameson
,
K. A.
Hogquist
.
2011
.
Alternative memory in the CD8 T cell lineage.
Trends Immunol.
32
:
50
56
.
18
Weinreich
,
M. A.
,
O. A.
Odumade
,
S. C.
Jameson
,
K. A.
Hogquist
.
2010
.
T cells expressing the transcription factor PLZF regulate the development of memory-like CD8+ T cells.
Nat. Immunol.
11
:
709
716
.
19
Kurzweil
,
V.
,
A.
LaRoche
,
P. M.
Oliver
.
2014
.
Increased peripheral IL-4 leads to an expanded virtual memory CD8+ population.
J. Immunol.
192
:
5643
5651
.
20
Lee
,
Y. J.
,
K. L.
Holzapfel
,
J.
Zhu
,
S. C.
Jameson
,
K. A.
Hogquist
.
2013
.
Steady-state production of IL-4 modulates immunity in mouse strains and is determined by lineage diversity of iNKT cells.
Nat. Immunol.
14
:
1146
1154
.
21
Akue
,
A. D.
,
J.-Y.
Lee
,
S. C.
Jameson
.
2012
.
Derivation and maintenance of virtual memory CD8 T cells.
J. Immunol.
188
:
2516
2523
.
22
Sosinowski
,
T.
,
J. T.
White
,
E. W.
Cross
,
C.
Haluszczak
,
P.
Marrack
,
L.
Gapin
,
R. M.
Kedl
.
2013
.
CD8α+ dendritic cell trans presentation of IL-15 to naive CD8+ T cells produces antigen-inexperienced T cells in the periphery with memory phenotype and function.
J. Immunol.
190
:
1936
1947
.
23
Le Campion
,
A.
,
C.
Bourgeois
,
F.
Lambolez
,
B.
Martin
,
S.
Léaument
,
N.
Dautigny
,
C.
Tanchot
,
C.
Pénit
,
B.
Lucas
.
2002
.
Naive T cells proliferate strongly in neonatal mice in response to self-peptide/self-MHC complexes.
Proc. Natl. Acad. Sci. USA
99
:
4538
4543
.
24
Schüler
,
T.
,
G. J.
Hämmerling
,
B.
Arnold
.
2004
.
Cutting edge: IL-7-dependent homeostatic proliferation of CD8+ T cells in neonatal mice allows the generation of long-lived natural memory T cells.
J. Immunol.
172
:
15
19
.
25
Morris
,
S. C.
,
S. M.
Heidorn
,
D. R.
Herbert
,
C.
Perkins
,
D. A.
Hildeman
,
M. V.
Khodoun
,
F. D.
Finkelman
.
2009
.
Endogenously produced IL-4 nonredundantly stimulates CD8+ T cell proliferation.
J. Immunol.
182
:
1429
1438
.
26
Haluszczak
,
C.
,
A. D.
Akue
,
S. E.
Hamilton
,
L. D. S.
Johnson
,
L.
Pujanauski
,
L.
Teodorovic
,
S. C.
Jameson
,
R. M.
Kedl
.
2009
.
The antigen-specific CD8+ T cell repertoire in unimmunized mice includes memory phenotype cells bearing markers of homeostatic expansion.
J. Exp. Med.
206
:
435
448
.
27
Lee
,
J.-Y.
,
S. E.
Hamilton
,
A. D.
Akue
,
K. A.
Hogquist
,
S. C.
Jameson
.
2013
.
Virtual memory CD8 T cells display unique functional properties.
Proc. Natl. Acad. Sci. USA
110
:
13498
13503
.
28
Rabinovich
,
B. A.
,
J.
Li
,
J.
Shannon
,
R.
Hurren
,
J.
Chalupny
,
D.
Cosman
,
R. G.
Miller
.
2003
.
Activated, but not resting, T cells can be recognized and killed by syngeneic NK cells.
J. Immunol.
170
:
3572
3576
.
29
Ventre
,
E.
,
L.
Brinza
,
S.
Schicklin
,
J.
Mafille
,
C.-A.
Coupet
,
A.
Marçais
,
S.
Djebali
,
V.
Jubin
,
T.
Walzer
,
J.
Marvel
.
2012
.
Negative regulation of NKG2D expression by IL-4 in memory CD8 T cells.
J. Immunol.
189
:
3480
3489
.
30
Jubin
,
V.
,
E.
Ventre
,
Y.
Leverrier
,
S.
Djebali
,
K.
Mayol
,
M.
Tomkowiak
,
J.
Mafille
,
M.
Teixeira
,
D. Y.-L.
Teoh
,
B.
Lina
, et al
.
2012
.
T inflammatory memory CD8 T cells participate to antiviral response and generate secondary memory cells with an advantage in XCL1 production.
Immunol. Res.
52
:
284
293
.
31
Anderson
,
K. G.
,
K.
Mayer-Barber
,
H.
Sung
,
L.
Beura
,
B. R.
James
,
J. J.
Taylor
,
L.
Qunaj
,
T. S.
Griffith
,
V.
Vezys
,
D. L.
Barber
,
D.
Masopust
.
2014
.
Intravascular staining for discrimination of vascular and tissue leukocytes.
Nat. Protoc.
9
:
209
222
.
32
Silver
,
J. D.
,
M. E.
Ritchie
,
G. K.
Smyth
.
2009
.
Microarray background correction: maximum likelihood estimation for the normal-exponential convolution.
Biostatistics
10
:
352
363
.
33
Huber
,
W.
,
A.
von Heydebreck
,
H.
Sültmann
,
A.
Poustka
,
M.
Vingron
.
2002
.
Variance stabilization applied to microarray data calibration and to the quantification of differential expression.
Bioinformatics
18
(
Suppl. 1
):
S96
S104
.
34
Bolstad
,
B. M.
,
R. A.
Irizarry
,
M.
Astrand
,
T. P.
Speed
.
2003
.
A comparison of normalization methods for high density oligonucleotide array data based on variance and bias.
Bioinformatics
19
:
185
193
.
35
Irizarry
,
R. A.
,
B. M.
Bolstad
,
F.
Collin
,
L. M.
Cope
,
B.
Hobbs
,
T. P.
Speed
.
2003
.
Summaries of affymetrix geneChip probe level data.
Nucleic Acids Res.
31
:
e15
.
36
Talloen
,
W.
,
D.-A.
Clevert
,
S.
Hochreiter
,
D.
Amaratunga
,
L.
Bijnens
,
S.
Kass
,
H. W. H.
Göhlmann
.
2007
.
I/NI-calls for the exclusion of non-informative genes: a highly effective filtering tool for microarray data.
Bioinformatics
23
:
2897
2902
.
37
Moutaftsi
,
M.
,
B.
Peters
,
V.
Pasquetto
,
D. C.
Tscharke
,
J.
Sidney
,
H.-H.
Bui
,
H.
Grey
,
A.
Sette
.
2006
.
A consensus epitope prediction approach identifies the breadth of murine T(CD8+)-cell responses to vaccinia virus.
Nat. Biotechnol.
24
:
817
819
.
38
Soudja
,
S. M.
,
A. L.
Ruiz
,
J. C.
Marie
,
G.
Lauvau
.
2012
.
Inflammatory monocytes activate memory CD8(+) T and innate NK lymphocytes independent of cognate antigen during microbial pathogen invasion.
Immunity
37
:
549
562
.
39
Kohlmeier
,
J. E.
,
W. W.
Reiley
,
G.
Perona-Wright
,
M. L.
Freeman
,
E. J.
Yager
,
L. M.
Connor
,
E. L.
Brincks
,
T.
Cookenham
,
A. D.
Roberts
,
C. E.
Burkum
, et al
.
2011
.
Inflammatory chemokine receptors regulate CD8(+) T cell contraction and memory generation following infection.
J. Exp. Med.
208
:
1621
1634
.
40
Sung
,
J. H.
,
H.
Zhang
,
E. A.
Moseman
,
D.
Alvarez
,
M.
Iannacone
,
S. E.
Henrickson
,
J. C.
de la Torre
,
J. R.
Groom
,
A. D.
Luster
,
U. H.
von Andrian
.
2012
.
Chemokine guidance of central memory T cells is critical for antiviral recall responses in lymph nodes.
Cell
150
:
1249
1263
.
41
Slütter
,
B.
,
L. L.
Pewe
,
S. M.
Kaech
,
J. T.
Harty
.
2013
.
Lung airway-surveilling CXCR3(hi) memory CD8(+) T cells are critical for protection against influenza A virus.
Immunity
39
:
939
948
.
42
McNally
,
B.
,
M.
Willette
,
F.
Ye
,
S.
Partida-Sanchez
,
E.
Flaño
.
2012
.
Intranasal administration of dsRNA analog poly(I:C) induces interferon-α receptor-dependent accumulation of antigen experienced T cells in the airways.
PLoS One
7
:
e51351
.
43
Ely
,
K. H.
,
L. S.
Cauley
,
A. D.
Roberts
,
J. W.
Brennan
,
T.
Cookenham
,
D. L.
Woodland
.
2003
.
Nonspecific recruitment of memory CD8+ T cells to the lung airways during respiratory virus infections.
J. Immunol.
170
:
1423
1429
.
44
Ely
,
K. H.
,
T.
Cookenham
,
A. D.
Roberts
,
D. L.
Woodland
.
2006
.
Memory T cell populations in the lung airways are maintained by continual recruitment.
J. Immunol.
176
:
537
543
.
45
Nourshargh
,
S.
,
R.
Alon
.
2014
.
Leukocyte migration into inflamed tissues.
Immunity
41
:
694
707
.
46
White
,
J. T.
,
E. W.
Cross
,
M. A.
Burchill
,
T.
Danhorn
,
M. D.
McCarter
,
H. R.
Rosen
,
B.
O’Connor
,
R. M.
Kedl
.
2016
.
Virtual memory T cells develop and mediate bystander protective immunity in an IL-15-dependent manner.
Nat. Commun.
7
:
11291
.
47
Goldrath
,
A. W.
,
C. J.
Luckey
,
R.
Park
,
C.
Benoist
,
D.
Mathis
.
2004
.
The molecular program induced in T cells undergoing homeostatic proliferation.
Proc. Natl. Acad. Sci. USA
101
:
16885
16890
.
48
Kaech
,
S. M.
,
W.
Cui
.
2012
.
Transcriptional control of effector and memory CD8+ T cell differentiation.
Nat. Rev. Immunol.
12
:
749
761
.
49
Hikono
,
H.
,
J. E.
Kohlmeier
,
K. H.
Ely
,
I.
Scott
,
A. D.
Roberts
,
M. A.
Blackman
,
D. L.
Woodland
.
2006
.
T-cell memory and recall responses to respiratory virus infections.
Immunol. Rev.
211
:
119
132
.
50
Ray
,
S. J.
,
S. N.
Franki
,
R. H.
Pierce
,
S.
Dimitrova
,
V.
Koteliansky
,
A. G.
Sprague
,
P. C.
Doherty
,
A. R.
de Fougerolles
,
D. J.
Topham
.
2004
.
The collagen binding alpha1beta1 integrin VLA-1 regulates CD8 T cell-mediated immune protection against heterologous influenza infection.
Immunity
20
:
167
179
.
51
Woodland
,
D. L.
,
J. E.
Kohlmeier
.
2009
.
Migration, maintenance and recall of memory T cells in peripheral tissues.
Nat. Rev. Immunol.
9
:
153
161
.
52
Lee
,
A.
,
S. P.
Park
,
C. H.
Park
,
B. H.
Kang
,
S. H.
Park
,
S.-J.
Ha
,
K. C.
Jung
.
2015
.
IL-4 induced innate CD8+ T cells control persistent viral infection.
PLoS Pathog.
11
:
e1005193
.
53
Hugues
,
S.
,
A.
Scholer
,
A.
Boissonnas
,
A.
Nussbaum
,
C.
Combadière
,
S.
Amigorena
,
L.
Fetler
.
2007
.
Dynamic imaging of chemokine-dependent CD8+ T cell help for CD8+ T cell responses.
Nat. Immunol.
8
:
921
930
.
54
Kieper
,
W. C.
,
J. T.
Burghardt
,
C. D.
Surh
.
2004
.
A role for TCR affinity in regulating naive T cell homeostasis.
J. Immunol.
172
:
40
44
.
55
Fulton
,
R. B.
,
S. E.
Hamilton
,
Y.
Xing
,
J. A.
Best
,
A. W.
Goldrath
,
K. A.
Hogquist
,
S. C.
Jameson
.
2015
.
The TCR’s sensitivity to self peptide-MHC dictates the ability of naive CD8(+) T cells to respond to foreign antigens.
Nat. Immunol.
16
:
107
117
.
56
King
,
C.
,
A.
Ilic
,
K.
Koelsch
,
N.
Sarvetnick
.
2004
.
Homeostatic expansion of T cells during immune insufficiency generates autoimmunity.
Cell
117
:
265
277
.

The authors have no financial conflicts of interest.

Supplementary data